How well does bindfs behave, when the user makes some modification to the lower directory directly (i.e. not through bindfs) ?
(Motivation: I assumed that it would not be a good idea to wrap a directory with bindfs, at a time when that directory was being written to. But it made me curious.
In theory I'd also be interested if there was an answer for another FUSE filesystem or one based on general FUSE principles, but I can't think of a good way to ask that here).
Kernel bind mounts -
mount --bind - share the VFS cache in some way. That is an example where concurrent modifications are fully supported, as far as I know.
An example where modifying a backing filesystem is not supported at all is the kernel
Changes to the underlying filesystems while part of a mounted overlay filesystem are not allowed. If the underlying filesystem is changed, the behavior of the overlay is undefined, though it will not result in a crash or deadlock.